letters to nature
and is uncorrelated from one day to the next. This gives a standard
error in the trend estimate of 0.08% per decade20. An additional
error is the uncertainty in determining the intersatellite biases,
which are estimated from data collected during periods of satellite
overlap (1±4 years). Monte Carlo simulations show that the
uncertainty in specifying the intersatellite biases introduces a
0.2% per decade standard error in the trend estimate. These
estimates of trend error do not include the additional uncertainty
involved with questions of prediction or representation (that is, how
well the 1987±1998 period represents longer-term climate change).
As a check on the error analysis, we divide the SSM/I data into two
sets: local morning and local evening. For each set, completely
independent derivations of the decadal trends for the three latitude
zones give a maximum difference between the morning and evening
trends of 0.2% per decade. This agreement indicates that diurnal
effects only weakly in¯uence the SSM/I data, in marked contrast to
the signi®cant diurnal problems affecting the MSUs. A complicated
procedure is required to remove the diurnal effect from the MSU
observations10, and this possibly introduces spurious trends.
The retrieval of W from SSM/I observations is less problematic
than the TS and TA retrievals in other ways as well. The 22-GHz
radiance observed by SSM/I is directly related to W, whereas the
MSU measures the middle-to-upper troposphere and then infers
the lower-tropospheric temperature21. Furthermore, assuming a
constant Hrel, the 22-GHz water-vapour radiance is three times
more sensitive to changes in air temperature than MSU 54-GHz
radiance. For AVHRR, a continuous calibration with in situ data is
required to remove the cooling effect of atmospheric aerosols such
as those emitted by volcanic eruptions22 and to correct for instrument drift and intersatellite biases2. The robustness and accuracy of
the SSM/I W retrieval makes it a useful validation tool for the more
complex retrievals of TA and TS, although we must keep in mind that
they are different physical variables.
The consistency we now see emerging from various satellite
observations provides new information on climate dynamics and
should help resolve some of the past controversies concerning errors
in the satellite data. It is notable that a relatively simple constant-Hrel
plus MALR model closely predicts the observed interannual and
decadal variations of W, TS and TA when zonally averaged over the
tropics. Furthermore, the evidence here shows that the marine
atmosphere has signi®cantly warmed and moistened over the past
decade.
Received 8 March; accepted 26 November 1999.
1. Spencer, R. W. & Christy, J. R. Precise monitoring of global temperature trends from satellites. Science
247, 1558±1562 (1990).
2. Reynolds, R. W. & Smith, T. M. Improved global sea surface temperature analyses using optimum
interpolation. J. Clim. 7, 929±948 (1994).
3. Wentz, F. J. A well-calibrated ocean algorithm for special sensor microwave/imager. J. Geophys. Res.
102, 8703±8718 (1997).
4. Hurrell, J. W. & Trenberth, K. E. Spurious trends in satellite MSU temperatures from merging different
satellite records. Nature 386, 164±167 (1997).
5. Hurrell, J. W. & Trenberth, K. E. Dif®culties in obtaining reliable temperature trends: reconciling the
surface and satellite microwave sounding unit records. J. Clim. 11, 945±967 (1998).
6. Wentz, F. J. & Schabel, M. Effects of satellite orbital decay on MSU lower tropospheric temperature
trends. Nature 394, 661±664 (1998).
7. Stephens, G. L. On the relationship between water vapor over the oceans and sea surface temperature.
J. Clim. 3, 634±645 (1990).
8. Jackson, D. L. & Stephens, G. L. A study of SSM/I-derived columnar water vapor over the global
oceans. J. Clim. 8, 2025±2038 (1995).
9. Randel, D. L. et al. A new global water vapor dataset. Bull. Am. Meteorol. Soc. 77, 1233±1246 (1996).
10. Christy, J. R., Spencer, R. W. & Lobl, E. S. Analysis of the merging procedure for the MSU daily
temperature time series. J. Clim. 11, 2016±2041 (1998).
11. Bony, S., Duvel, J. -P. & Le Treut, H. Observed dependence of the water vapor and clear-sky
greenhouse effect on sea surface temperature: comparison with climate warming experiments. Clim.
Dyn. 11, 307±320 (1995).
12. Stone, S. H. & Carlson, J. H. Atmospheric lapse rate regimes and their parameterization. J. Atmos. Sci.
36, 415±423 (1979).
13. Peixoto, J. P. & Oort, A. H. Physics of Climate (American Institute of Physics Press, New York, 1991).
14. Gaffen, D. J., Elliott, W. P. & Robock, A. Relationships between tropospheric water vapor and surface
temperature as observed by radiosondes. Geophys. Res. Lett. 19/18, 1839±1842 (1992).
15. Ross, R. J. & Elliott, W. P. Tropospheric water vapor climatology and trends over north America:
1973±93. J. Clim. 9±12, 3561±3574 (1996).
416
16. Zhai, P. & Eskridge, R. E. Atmospheric water vapor over China. J. Clim. 10, 2643±2652 (1997).
17. Gutzler, D. Low-frequency ocean±atmosphere variability across the tropical western Paci®c. J. Atmos.
Sci. 53, 2773±2785 (1996).
18. Schroeder, S. R. & McGuirk, J. P. Widespread tropical atmospheric drying from 1979 to 1995. Geophys.
Res. Lett. 25±9, 1301±1304 (1998).
19. Ross, R. J. & Gaffen, D. J. Comment on Widespread tropical atmospheric drying from 1979 to 1995, by
Schroeder and McGuirk. Geophys. Res. Lett. 25±23, 4357±4358 (1998).
20. Wilks, D. S. Statistical Methods in the Atmospheric Sciences (Academic, New York, 1995).
21. Spencer, R. W. & Christy, J. R. Precision and radiosonde validation of satellite gridpoint temperature
anomalies. Part II: A tropospheric retrieval and trends during 1979±1990. J. Clim. 53, 858±866
(1992).
22. Reynolds, R. W. Impact of Mount Pinatubo aerosols on satellite-derived sea surface temperatures. J.
Clim. 6, 768±775 (1993).
Acknowledgements
This work was supported by NASA as part of their path®nder Data Set program.
Correspondence and requests for materials should be addressed to F.J.W.
(e-mail: [email protected]).
.................................................................
The Pleistocene serpent Wonambi
and the early evolution of snakes
John D. Scanlon*² & Michael S. Y. Lee*
* Department of Zoology, University of Queensland, Brisbane QLD 4072,
Australia
² Department of Biological Sciences, University of New South Wales, Sydney,
NSW 2052, Australia
..............................................................................................................................................
The Madtsoiidae were medium sized to gigantic snakes with a
fossil record extending from the mid-Cretaceous to the Pleistocene, and spanning Europe, Africa, Madagascar, South America
and Australia1±3. This widely distributed group survived for about
90 million years (70% of known ophidian history), and potentially
provides important insights into the origin and early evolution of
snakes. However, madtsoiids are known mostly from their vertebrae, and their skull morphology and phylogenetic af®nities
have been enigmatic. Here we report new Australian material of
Wonambi, one of the last-surviving madtsoiids4±6, that allows the
®rst detailed assessment of madtsoiid cranial anatomy and relationships. Despite its recent age, which could have overlapped
with human history in Australia, Wonambi is one of the most
primitive snakes knownÐas basal as the Cretaceous forms
Pachyrhachis7 and Dinilysia8. None of these three primitive
snake lineages shows features associated with burrowing, nor do
any of the nearest lizard relatives of snakes (varanoids). These
phylogenetic conclusions contradict the widely held `subterranean' theory of snake origins9±12, and instead imply that burrowing snakes (scolecophidians and anilioids) acquired their fossorial
adaptations after the evolution of the snake body form and
jaw apparatus in a large aquatic or (surface-active) terrestrial
ancestor.
Serpentes Linnaeus, 1758
Madtsoiidae Hoffstetter, 1961
Wonambi Smith, 1976
Revised diagnosis. Neural spines of vertebrae high, sloping, posterodorsally, with sharp-edged anterior lamina extending to near
anterior edge of zygosphene; transverse processes extending laterally
beyond zygaphophyses in most trunk vertebrae, synapophyses with
concave dorsal edge in lateral view; zygosphene relatively narrow,
with steep facets (20±308 from vertical); zygapophyses inclined 208
or more above horizontal; haemal keel in middle and posterior
trunk region narrow and weakly de®ned laterally, but often distinctly bi®d or tri®d on the posterior third of the centrum.
© 2000 Macmillan Magazines Ltd
NATURE | VOL 403 | 27 JANUARY 2000 | www.nature.com
letters to nature
Figure 1 Reconstruction of the skull of Wonambi naracoortensis Smith, 1976. a, Dorsal,
b, palatal and c, lateral views. Small outline drawings indicate parts known in
W. naracoortensis (stipple) and in W. barriei (cross-hatching). Scale bar, 30 mm (based on
the largest known individual at type locality, FU1762). bo, basioccipitial; ec, ectopterygoid;
cor, coronoid; de, dentary; fr, frontal; ju, jugal; max, maxilla; na, nasal; op, opisthotic; pal,
palatine; pa, parietal; pm, premaxilla; prf, prefrontal; pob, postorbital; pof, postfrontal;
pro, prootic; pt, pterygoid; qa, quadrate; sm, septomaxilla; so, supraoccipital; sph,
sphenoid; sur, surangular; sut, supratemporal; vo, vomer.
Pterygoid tooth row near middle of bone, away from medial edge,
and basipterygoid facet narrow and facing medially as much as
dorsally; ectopterygoid process of pterygoid triangular in palatal
view. Maxilla and dentary relatively elongate and depressed; maxilla
with deep, anterolaterally directed trough on suborbital surface.
Wonambi naracoortensis Smith, 1976
Diagnosis. Large size (trunk vertebrae more than 20 mm and often
greater than 30 mm wide, mandible length up to 160 mm, total
length estimated to exceed 5 m); relatively small neural canal and
deep zygosphere; zygapophyses inclined about 258 above horizontal;
parazygantral foramina often small and multiple, weakly distinguished from pits in the same region. Palatine broad (mediolateral
dimension of choanal process greater than that of tooth-bearing
portion of palatine).
Material. Abundant vertebrae (including holotype mid-trunk vertebra, South Australian Museum P16168), ribs, and the following
elements of the skull: complete maxillae, prootic, exoccipitalopisthotics, basioccipital, dentaries, compound mandibular elements; near-complete sphenoid, frontal, palatine, ectopterygoid,
and parietal; and fragment of pterygoid. Most of these are represented in the partial skeletons from Henschke's Cave, Naracoorte
(SAM P30178)5, but some have not been identi®ed until this study.
Range. Pliocene and Pleistocene, southern Australia (Western
Australia, South Australia, New South Wales)6.
Wonambi barriei Scanlon, sp. nov.
Etymology. Named in honour of D. John Barrie, who has collected
and prepared most of the material of W. naracoortensis.
Diagnosis. Trunk vertebrae less than 15 mm wide in adults (total
length estimated to be less than 3 m); relatively large neural canal
and shallow zygosphene; zygapophyses approximately 208 above
horizontal. Single, large parazygantral foramen on each side of
neural arch. Palatine relatively narrow (mediolateral dimension of
choanal process subequal to that of tooth-bearing portion).
Material. Nine well-preserved vertebrae from WW Site, Riversleigh
(northwest Queensland, Australia) represent most body regions,
probably from a single individual (Queensland Museum F23038,
holotype mid-trunk vertebra, and F40194, paratypes). Abundant
postcranial material from various other Oligo-Miocene sites at
Riversleigh is referred to this taxon, including a series of cloacal
and anterior caudal vertebrae from CS Site (QM F40189). Referred
cranial remains include associated jaw elements from CS Site (nearcomplete left palatine QM F40190 and fragments of the right
palatine QM F40191, right and left pterygoids QM F23047 and
23048, articular and prearticular region of left mandible QM
F23077, 23078), and fragments of two maxillae from CS (QM
F39932) and WW Site (QM F40193).
Range. Late Oligocene to Early Miocene of Riversleigh, northwest
Queensland13.
Comments. Elements of W. barriei are half the linear dimensions of
corresponding elements of W. naracoortensis, implying a roughly
eightfold difference in mass. Vertebrae of W. barriei can be interpreted as coming from adults of a small form because of the full
extent of perichondral ossi®cation, and the proportions of the
centrum and neural arch. The two species are similar in these
NATURE | VOL 403 | 27 JANUARY 2000 | www.nature.com
© 2000 Macmillan Magazines Ltd
417
letters to nature
proportions, whereas if the smaller forms merely represented
juveniles of W. naracoortensis they would have shorter, broader
and shallower centra, and dorsoventrally thinner neural arches15.
W. barriei exhibits narrow palatine bones, an autapomorphy not
found in W. naracoortensis or other madtsoiids14. It also lacks the
autapomorphy of W. naracoortensis, small and numerous parazygantral foramina. Neither of these traits appears to be correlated
with size in madtsoiids or other snakes. Each species spans a
considerable stratigraphic interval without noticeable change in
size or morphology.
The anatomy of Wonambi is revised here on the basis of newly
identi®ed elements (for example, frontal and ectopterygoid of
W. naracoortensis and pterygoid of W. barriei) and a re-interpretation of known elements, which have only been brie¯y discussed and
not yet studied phylogenetically5. No other madtsoiid taxon is
represented by such complete material; therefore, Wonambi provides the best available evidence for the phylogenetic relationships
of madtsoiids with other snakes. The known elements of other
madtsoiids1,3,14 are similar to those of Wonambi and suggest that
madtsoiids remained morphologically conservative throughout
their long history.
The skull of Wonambi (Figs 1 and 2a±c) is long and low, the snout
rounded, the palate very broad, and the braincase constricted
behind the orbits but widening posteriorly. The parietal and frontal
are most like those of Dinilysia8 in shape (including the large lateral
crests of the parietal5) and attachments to each other and surrounding bones. The snout is rigidly attached to the frontal (Fig. 1) at a
vertical frontonasal suture; the frontonasal (prokinetic) joint of
modern snakes is thus poorly developed (condition unknown in
Pachyrhachis7, in Dinilysia the prokinetic joint is described as having
only limited mobility8,16). The upper jaws have relatively tight and
extensive connections to the central elements of the skull. The ¯at,
pitted anterior end of the maxilla suggests a short ligamentous
connection between the maxilla and premaxilla. The prefontal is
sutured to the maxilla but movably articulated with the frontal.
There is an extensive palatine±vomer contact and small but
distinct facets on the pterygoids for the basipterygoid articulations.
A facet-like `trough' on the suborbital region of the maxilla
suggests retention of a jugal14 which was also present in
Pachyrhachis7 and possibly Dinilysia8 but lost in all extant snakes.
These primitive features indicate that kinesis of the upper jaws was
more limited than in modern snakes, but similar to that of
Figure 2 Selected elements of Wonambi exhibiting phylogenetically important characters.
a, External view of braincase of Wonambi naracoortensis (SAMP30178) in left lateral view.
Scale bar, 10 mm. b, Internal view of braincase in right lateral and slightly dorsal view.
c, Articular region of right mandible of W. barriei (QMF23077) showing sutures between
articular, surangular and prearticular. Scale bar, 3 mm. d, Reconstructed tooth, with part
of base cut away, showing plicidentine on internal and external surfaces of tooth base
(based on FU1762). Scale bar, 2 mm. e, Holotype mid-dorsal vertebra of W. barriei
(QMF23038) in lateral, anterior and posterior views. Scale bar, 5 mm. f, Caudal vertebra
of W. barriei (QMF40189) in lateral and ventral views, showing paired pedicels near
posterior end of centrum. Scale bar, 4 mm. g, Partial caudal vertebra of W. naracoortensis
(SAMP30178) in lateral and anterior views showing articulated chevron. Scale bar, 5 mm.
h, Head of left rib of W. naracoortensis (SAMP30178) in anterolateral view showing dorsal
tubercle. Scale bar, 10 mm. Abbreviations as in Fig. 1 plus: art, articular; bpt process (l),
left basipterygoid process; bpt process (r), right basipterygoid process; cen, centrum;
chev, chevron; cri cir, crista circumfenestralis; dor sel, dorsum sellae; dor tub, dorsal
tubercle; fen ov, fenestra ovalis; fen pseu, fenestra pseudorotunda; ns, neural spine; par
for, paracotylar foramen; par proc, paroccipital process; ped, pedicel for chevron; pra,
prearticular; pzf, parazygantral foramen; top, transverse process; vid can (ant), anterior
opening of vidian canal; vid can (post), posterior opening of vidian canal; V2+3, trigeminal
foramen.
418
© 2000 Macmillan Magazines Ltd
NATURE | VOL 403 | 27 JANUARY 2000 | www.nature.com
letters to nature
proximal facets slightly constricted between the strongly concave
diapophyseal and ¯atter parapophyseal surfaces, and there is a
prominent dorsal process (tuber costae) adjacent to the facet (Fig.
2h). No limb or girdle elements are yet known.
Madtsoiids are usually interpreted as an early appearing group of
advanced snakes related to boas and pythons1,4,5,19, although it has
been suggested that they might represent a much more primitive
lineage2,3. Until recently, most fossil snake `families' (including
madtsoiids) were not represented by suf®cient cranial material for
rigorous assessment of relationships1±3. The new information on
Wonambi allows madtsoiids to be included in such an analysis for
the ®rst time. Accordingly, a phylogenetic analysis was performed
including all living and fossil snake families (except those known
from only postcranial elements), and incorporating the most
extensive set of morphological characters so far assembled for
snakes (see Methods). Very similar trees resulted whether multistate
characters were ordered or unordered, and whether varanoid
lizards, or amphisbaenians and dibamids, were used to infer
character polarity. The results (Fig. 3) show that madtsoiids are
among the most `primitive' of snakes: as basal as the upper
Scolecophidia
27,100
Xenopeltidae Booidea
9,98
8,97
a
s
de
du
or
br
oi
lu
ch
5,85
Co
ro
Ac
Bo
lye
riid
ph
lio
ga
Un
ae
iin
ae
in
hi
ae
op
in
id
on
Tr
op
th
Boidae
3,83
2,40
Anilioidea
2,42
Py
Bo
in
ae
ae
in
yc
Er
Lo
xo
ce
m
ltis
ae
Xe
no
pe
id
hi
elt
Ur
op
us
s
lu
s
hi
oc
om
lin
Cy
1,45
An
s
dr
op
pi
ale
iliu
An
om
An
18,100
ae
e
da
di
id
ae
op
id
op
ot
hl
Ty
p
pt
Le
+D
i
ni
lys
yp
ia
hl
id
oi
ts
ad
+M
+P
ac
hy
op
hi
id
ae
ae
ae
anilioids17. However, the mandible was longer than the skull, and
the supratemporals and/or quadrates must have projected posteriorly, as in Pachyrhachis and macrostomatan snakes, facilitating the
swallowing of large items.
The anterior trunk region, representing about one third of all
trunk vertebrae, comprises relatively high and narrow vertebrae
with prominent single hypapophyses, and synapophyses projecting
partly below the centrum. Mid-trunk vertebrae are broader, with
slightly lower and longer neural spines, more dorsally placed
synapophyses, and weak, posteriorly bi®d or tri®d haemal keels
(Fig. 2e). More posteriorly the trunk vertebrae become smaller but
relatively longer. Haemal keels disappear just before the cloacal
region, but reappear on the cloacals and caudals. There are at least
three cloacals with fused, forked lymphapophyses. At least one
anterior caudal (`pygal') lacks haemapophyses, but other caudals
have paired oval facets for their attachment on the posterior part of
the centrum, near the ventral midline (Fig. 2f). The haemapophyses, which remain articulated to the centrum of one caudal
vertebra in W. naracoortensis, are fused distally into a true chevron
(Fig. 2g). Trunk ribs are curved throughout their length, with their
Caenophidia
8,98
1,54
7,90
12,99
Macrostomata
10,98
C1-10: Alethinophidia
2,62
B1-10: Modern snakes
10,100
A1-6: Unnamed clade
6,93
Snakes
Figure 3 Cladogram (strict consensus of two most parsimonious trees, each with
length ˆ 649, consistency index ˆ 0:49, retention index ˆ 0:66) showing relationships among snake lineages and the very basal position of madtsoiids. This is based on a
phylogenetic analysis of 234 morphological characters in 20 snake taxa (see Methods).
Extinct taxa are denoted by `+'. The ®rst number at each clade refers to Bremer support;
the second refers to bootstrap frequency. Clade A, Pachyrhachis retains the following
primitive characters shared with varanoid lizards (the derived state that unites madtsoiids,
Dinilysia and modern snakes is listed afterwards in parentheses). A1, Exoccipitals not in
contact above foramen magnum (Fig. 1a) (exoccipitals in contact, excluding supraoccipital
from skull margin). A2, Parietal less than 40% of skull length, that is, premaxilla±occiput
dimension (Fig. 1a) (more than 40%). A3, Sagittal dimension of supraoccipital long, more
than 50% transverse dimension (Fig. 1a) (short, less than 50% transverse dimension). A4,
Coronoid overlaps lateral surface of surangular (Fig. 2c) (does not overlap lateral surface).
A5, Free forked cloacal ribs, that is, `lymphapophyses', absent (present). A6, Rib heads
without dorsal process (Fig. 2h) (process present). In addition, Pachyrhachis is more
primitive than all living snakes in retaining a jugal bone, tibia, ®bula and tarsals, in having
a distinct cervical (neck) region with short narrow ribs, and in possessing fewer than 120
precloacal vertebrae7,18 (these primitive features might also characterize madtsoiids and/
or Dinilysia, but the relevant regions in both are insuf®ciently known). Clade B,
Pachyrhachis (where known), madtsoiids and Dinilysia retain the following primitive
characters shared with varanoid lizards (the derived state that unites scolecophidian and
alethinophidian snakes is listed afterwards in parentheses). B1, Ectopterygoid clasps
dorsal and ventral surfaces of pterygoid (ectopterygoid overlaps ventral surface of
pterygoid only). B2, Basipterygoid process prominent (Fig. 2) (basipterygoid processes
reduced or lost; re-acquired in some booids). B3, Alar process of prootic long, narrow and
distinct (Fig. 2) (alar process not distinct). B4, Crista circumfenestralis does not converge
NATURE | VOL 403 | 27 JANUARY 2000 | www.nature.com
to enclose stapedial footplate (Fig. 2a) (crista converges to partially or fully enclose
stapedial footplate). B5, Paroccipital process very long (Fig. 2a) (reduced or absent). B6,
Articular and surangular not fused in region of articular facet (Fig. 2c) (reduced or absent).
B6, Articular and surangular not fused in region of articular facet (Fig. 2c) (articular and
surangular fused). B7, Zygosphenal buttress with concave anterior edge between
zygosphenes (Fig. 2e) (straight). B8, Paracotylar foramina present (Fig. 2e) (absent;
reversals within Macrostomata). B9, Haemapophyses not fused to caudal centra (Fig. 2f, g)
(fused to centra when present). B10, Haemapophyses are chevron-shaped, that is, joined
distally (Fig. 2g) (separated distally when present). Madtsoiids are more primitive than
Dinilysia and modern snakes in retaining plicidentine on the tooth bases (Fig. 2d), and
lacking prezygapophysial processes (Fig. 2e); however, con¯icting characters mean that
the relationships between madtsoiids, Dinilysia and modern snakes remain unresolved.
Clade C, Pachyrhachis, madtsoiids and Dinilysia share with scolecophidians the following
primitive (varanoid-like) characters, which are modi®ed in alethinophidians (derived state
that unites alethinophidians is listed afterwards in parentheses). C1, Medial descending
lamina of frontal absent (present). C2, Frontoparietal suture straight or M-shaped (Fig. 1a)
(suture U-shaped and deeply concave anteriorly). C3, Vidian canal not opening
intracranially (Fig. 2a, b) (opening intracranially). C4, Laterosphenoid absent (Fig. 2a, b)
(laterosphenoid present). C5, Coronoid process formed entirely by coronoid (Fig. 1c)
(coronoid process formed partly or entirely by surangular). In addition, scolecophidians are
more primitive than alethinophidians in several soft anatomical traits which cannot be
scored in the fossil taxa. C6, Undifferentiated ventral scales (midventral scales expanded
transversely). C7, Presence of ®rst branchial arch elements (absent). C8, Adductor
mandibulae externus pars super®cialis inserts on adductor externus medialis (inserts on
adductor externus profundus). C9, Single pair of thymus glands (two pairs). C10, Nonlobed kidneys (lobed).
© 2000 Macmillan Magazines Ltd
419
letters to nature
Cretaceous Dinilysia8,20 and only slightly more derived than the
most basal known snake, the mid-Cretaceous Pachyrhachis7. In
particular, all three fossil taxa lie outside the clade consisting of
living snakes; this exclusion is supported by numerous compelling
characters, a Bremer index of 10 and a bootstrap value of 100%
(Fig. 3).
This analysis refutes the suggestion that Pachyrhachis is a derived
rather than primitive snake21. Uniting Pachyrhachis with
Macrostomata21 entails a signi®cant increase in tree length (28
steps, P , 0:0001; see Methods) and implies reacquisition of hindlimbs and several skull bones. Other positions have also been
suggested for madtsoiids (usually interpreted as booids1,4,5,19) and
Dinilysia (as an anilioid20, booid1 or basal alethinophidian12,22). All
these arrangements are signi®cantly less parsimonious than the
preferred tree (all entail between 10 and 26 extra steps and can be
rejected at P , 0:0001). The arrangement obtained for living taxa is
similar to those found in recent analyses1,12,22±24. There is strong
support for the monophyly of booids (boas, pythons and erycines),
boids (boas and pythons) and xenopeltids (Xenopeltis and
Loxocemus), clades which have been either poorly corroborated or
contradicted in recent studies12,22±24. There is also support for
monophyly12,22 rather than paraphyly23 of anilioids (Anilius,
Cylindrophis, uropeltids and Anomochilus). This region of the tree
is weak12,22,23, however, as con¯icting characters support nesting
scolecophidians within anilioids, as the sister group of Anomochilus.
Traits supporting this alternative arrangement, however, are almost
entirely losses and reductions associated with fossoriality and
miniaturization18. Within scolecophidians, a typhlopid±anomalepidid clade12,23 is strongly supported over a typhlopid±leptotyphlopid clade22.
The small, worm-like scolecophidians are usually thought to be
the most basal snakes1,11,12,22±24, an arrangement consistent with the
widely held idea that snakes evolved from tiny burrowing ancestors.
This analysis indeed con®rms that scolecophidians are the most
basal living snakes; however, three fossil taxa emerge as even more
basal than scolecophidians. Thus, the nearest relatives of snakes
(mosasaurs and terrestrial varanoid lizards) and the three most
basal snake lineages (Pachyrhachis, madtsoiids and Dinilysia) were
all large predators with wide gapes. Mosasaurs and Pachyrhachis
were marine, whereas madtsoiids and Dinilysia were either (surfaceactive) terrestrial or semi-aquatic1±6,14,16,25, showing none of the
extensive suite of traits correlated with fossoriality17. This challenges
the widespread assumption that snakes arose through a small,
burrowing ancestor9±12,23: taxa straddling the lizard±snake transition are all large predators and either marine or (surface-active)
terrestrial. These results suggest instead that the snake body form
arose in a large, non-fossorial ancestor, perhaps for eel-like
swimming26 or sliding through dense vegetation27,28. Fossoriality
makes its debut later in snake evolution, in `modern' forms such as
M
scolecophidians and anilioids.
Methods
Two-hundred and thirty-four morphological characters (191 osteological, 43 soft
anatomical), adapted and expanded from recent analyses of snake phylogeny7,23,29, were
scored for all snake `families'. Characters were polarized by reference to the nearest lizard
outgroups, mosasaurs and terrestrial varanoids18, and the tree rooted with a hypothetical
ancestor coded with these inferred primitive states. Use of an amphisbaenian±dibamid
clade as the outgroup (see ref. 18) yielded similar trees. Character descriptions and data
matrix in Nexus/PAUP format are available as Supplementary Information. Analyses were
performed using PAUP version 4.0 (ref. 30). Bremer support was determined using
converse constraints; bootstrap values are based on 1,000 heuristic replicates. Variability
within terminal taxa was treated as uncertainty (regarding the primitive state) when
calculating tree lengths. Two analyses were carried out: multistate characters ordered
according to morphoclines where possible, as discussed in the character descriptions; and
multistate characters all unordered. The branch-and-bound search found two most
parsimonious cladograms in the `ordered' analysis, with madtsoiids and Dinilysia forming
either a clade or successive sister taxa to living snakes; the strict consensus of these is shown
in Fig. 3. The `unordered' analysis found a single most parsimonious cladogram
corresponding to the ®rst topology. Thus, the phylogenetic results are not dependent on
assumptions of character state evolution (`ordering'). Only unambiguous characters
420
(those invariant under accelerated and delayed optimization) are listed in the clade
diagnoses in Fig. 3. To test alternative proposed positions for each fossil taxon, the
`backbone constraints' command was used: the fossil taxon in question was constrained to
form a clade with its putative living relatives to the exclusion of other living taxa, but other
fossil taxa were allowed to `¯oat'. The most parsimonious tree(s) consistent with the
backbone constraint was compared with the most parsimonious (unconstrained) trees
using the Templeton and Kishino±Hasegawa tests in PAUP. Tree statistics, Bremer/
bootstrap support, and Templeton/Kishino±Hasegawa test values reported (Fig. 3) are
those pertaining to the `ordered' analysis; those in the `unordered' analysis were almost
identical.
Received 1 June; accepted 16 November 1999.
1. Rage, J. C. Handbuch der PalaÈoherpetologie. Teil 11. Serpentes (Gustav Fischer, Stuttgart, 1984).
2. McDowell, S. B. in Snakes: Ecology and Evolutionary Biology (eds Seigel, R. A., Collins, J. T. C. & Novak,
S. S.) 1±50 (Macmillan, New York, 1987).
3. Rage, J. C. Fossil snakes from the Paleocene of SaÄo Jose de ItaboraõÂ, Brazil. Part I. Madtsoiidae,
Aniliidae. Palaeovertebrata 27, 109±144 (1998).
4. Smith, M. J. Small fossil vertebrates from Victoria Cave, Naracoorte, South Australia. IV. Reptiles.
Trans. R. Soc. S. Aust. 100, 39±51 (1976).
5. Barrie, D. J. Skull elements and associated remains of the Pleistocene boid snake Wonambi
naracoortensis. Mem. Qd Mus. 28, 139±151 (1990).
6. Scanlon, J. D. First records from Wellington Caves, New South Wales, of the extinct madtsoiid snake
Wonambi naracoortensis Smith, 1976. Proc. Linn. Soc. NSW 115, 233±238 (1995).
7. Lee, M. S. Y. & Caldwell, M. W. Anatomy and relationships of Pachyrhachis problematicus, a primitive
snake with hindlimbs. Phil. Trans. R. Soc. Lond. B 353, 1521±1552 (1998).
8. Estes, R., Frazzetta, T. H. & Williams, E. E. Studies on the fossil snake Dinilysia patagonica Woodward:
Part 1. Cranial morphology. Bull. Mus. Comp. Zool. Harv. 140, 25±74 (1970).
9. Walls, G. L. Ophthalmological implications for the early history of snakes. Copeia 1940, 1±8 (1940).
10. Bellairs, A. d'A. & Underwood, G. The origin of snakes. Biol. Rev. 26, 193±237 (1951).
11. Underwood, G. A Contribution to the Classi®cation of Snakes (British Museum (Natural History),
London, 1967).
12. Rieppel, O. A review of the origin of snakes. Evol. Biol. 22, 37±130 (1988).
13. Archer, M., Hand, S. J., Godthelp, H. & Creaser, P. in Actes du CongreÁs Biochrom '97 (eds. Aguilar, J.-P.,
Legendre, S. & Michaux, J.) 131±152 (EÂcole Pratique des Hautes EÂtudes Institute de Montpelier,
Montpelier, 1997).
14. Scanlon, J. D. Nanowana gen. nov., small madtsoiid snakes from the Miocene of Riversleigh:
sympatric species with divergently specialised dentition. Mem. Qd Mus. 41, 393±412 (1997).
15. LaDuke, T. C. The fossil snakes of Pit 91, Rancho La Brea, California. Nat. Hist. Mus. LA County
Contrib. Sci. 424, 1±28 (1991).
16. Frazzetta, T. H. Studies on the fossil snake Dinilysia patagonica Woodward. II. Jaw machinery in the
earliest snakes. Forma et Functio 3, 205±221 (1970).
17. Cundall, D. Feeding behaviour in Cylindrophis and its bearing on the evolution of alethinophidian
snakes. J. Zool. 237, 353±376 (1995).
18. Lee, M. S. Y. Convergent evolution and character correlation in burrowing reptiles: towards a
resolution of squamate phylogeny. Biol. J. Linn. Soc. 65, 369±453 (1998).
19. Underwood, G. in Morphology and Biology of Reptiles (eds Bellairs, A. d'A. & Cox, C. B.) 151±175
(Academic, London, 1976).
20. Woodward, A. S. On some extinct reptiles from Patagonia, of the genera Miolania, Dinilysia, and
Genyodectes. Proc. Zool. Soc. Lond. 1901, 169±184 (1901).
21. Zaher, H. The phylogenetic position of Pachyrhachis within snakes (Squamata, Lepidosauria). J. Vert.
Paleontol. 18, 1±3 (1998).
22. Kluge, A. G. Boine snake phylogeny and research cycles. Misc. Publ. Mus. Zool. Univ. Michigan 178,
1±58 (1991).
23. Cundall, D., Wallach, V. & Rossman, D. S. The systematic relationships of the snake genus
Anomochilus. Zool. J. Linn. Soc. 109, 275±299 (1993).
24. Heise, P. J., Maxson, L. R., Dowling, H. G. & Hedges, S. B. Higher-level snake phylogeny inferred from
mitochondrial DNA sequences of 12S rRNA and 16S rRNA genes. Mol. Biol. Evol. 12, 259±265 (1995).
25. Hecht, M. K. The vertebral morphology of the Cretaceous snake, Dinilysia patagonica Woodward. N.
Jb. Geol. PalaÈont. Mh. 1982, 523±532 (1982).
26. Nopcsa, F. Eidolosaurus und Pachyophis. Zwei neue Neocom-Reptilien. Palaeontographica 65, 99±154
(1923).
27. Cope, E. D. On the reptilian orders Pythonomorpha and Streptosauria. Proc. Boston Soc. Nat. Hist. 12,
250±266 (1869).
È ber Archaeophis proavus Mass., eine Schlange aus dem EocaÈn des Monte Bolca. Beitr. z.
28. Janensch, W. U
È sterreich-Ungarns 19, 1±33 (1906).
PalaÈont. Geol. O
29. Wallach, V. & Rainer, G. Visceral anatomy of the Malaysian snake genus Xenophidion, including a
cladistic analysis and allocation to a new family. Amphibia±Reptilia 19, 385±404 (1998).
30. Swofford, D. L. PAUP* Version 4ÐPhylogenetic Analysis Using Parsimony (*and Other Methods).
Computer program and documentation. (Sinauer, Sunderland, Massachusetts, 1999).
Supplementary information is available on Nature's World-Wide Web site (http://
www.nature.com) or as paper copy from the London editorial of®ce of Nature.
Acknowledgements
We thank D. J. Barrie, M. Archer, R. E. Molnar, N. Pledge and R. T. Wells for access to
materials, and V. Wallach, G. Underwood, J.-C. Rage, D. J. Barrie, S. E. Evans, H. W. Greene,
D. Cundall, M. W. Caldwell and A. G. Kluge for discussion. This research was supported by
Australian Research Council grants to M.L. and J.S. Work at Riversleigh was supported by
the Australian Research Council and University of New South Wales (to M. Archer), and
work at Naracoorte by Flinders University, the South Australian Museum, L. and G.
Henschke, the Barrie family and numerous volunteers.
Correspondence and requests for materials should be addressed to J.S.
(e-mail: jscanlon@ ultra.net.au).
© 2000 Macmillan Magazines Ltd
NATURE | VOL 403 | 27 JANUARY 2000 | www.nature.com